fix(ci): repair MySQL healthcheck so dependents wait for it properly

Exec-form healthcheck did not shell-expand $$MYSQL_USER and
$$MYSQL_PASSWORD, so mysqladmin received literal '$MYSQL_USER' as the
username. mysqladmin ping reports 'mysqld is alive' on connection
errors via stderr while still exiting non-zero with bad credentials,
making the check unstable and CI flakes (Adapter Tests
SharedTables/MySQL) appear with all tests failing on Connection
refused before MySQL was actually ready. Switch to CMD-SHELL so the
env vars expand at run time, ping over TCP on the configured port,
and use root/MYSQL_ROOT_PASSWORD which is always set.

Author: abnegate

docker-compose.yml

@@ -189,7 +189,7 @@ services:
     cap_add:
       - SYS_NICE
     healthcheck:
-      test: ["CMD", "mysqladmin", "ping", "-h", "localhost", "-u $$MYSQL_USER", "-p $$MYSQL_PASSWORD"]
+      test: ["CMD-SHELL", "mysqladmin ping -h 127.0.0.1 -P $$MYSQL_TCP_PORT -uroot -p$$MYSQL_ROOT_PASSWORD --silent"]
       interval: 10s
       timeout: 5s
       retries: 5
@@ -211,7 +211,7 @@ services:
     cap_add:
       - SYS_NICE
     healthcheck:
-      test: ["CMD", "mysqladmin", "ping", "-h", "localhost", "-u $$MYSQL_USER", "-p $$MYSQL_PASSWORD"]
+      test: ["CMD-SHELL", "mysqladmin ping -h 127.0.0.1 -P $$MYSQL_TCP_PORT -uroot -p$$MYSQL_ROOT_PASSWORD --silent"]
       interval: 10s
       timeout: 5s
       retries: 5